This invention relates to the field of optical communications, and more particularly, to an East-West separable Reconfigurable Optical Add/Drop Multiplexer (ROADM) for use in optical communications systems.
For several decades, fiber optics have been used for communication. Specifically, fiber optics are used for data transmission and other telecommunication applications. Despite the enormous information carrying capacity of fiber, as compared to conventional copper cable, the high cost of installing fiber optics presents a barrier to full implementation of fiber optics, particular as the “last mile”, from the central office to residences and businesses.
One common method used for optical communications is a synchronous optical network (SONET). SONET networks are often comprised of optical rings with interconnecting nodes attaching to the ring. In common SONET networks, a bi-directional line-switch ring is used for interconnection. There are two types of BLSRs: 2-fiber and 4-fiber. In a 4-fiber BLSR, the protection and working channels are over different physical fibers (lines). In a 2-fiber BLSR, half of the bandwidth on each of the two fibers is reserved for protection. In the 2-fiber BLSR, the protection mechanism used to protect against failures is the ring switch. In the 4-fiber BLSR, two protection mechanisms may be used to protect against failures: ring switch and span switch.
In a SONET network East/West Separability means that no single failure or repair can interrupt traffic to or from more than one direction. Hence all failures and repairs must either be protected through 1+1 equipment redundancy or have the same impact on traffic as a fiber or cable cut, which can be protected against by conventional 1:1 or 1+1 protection schemes. Add/Drop traffic failures may have the same impact as a single transmitter or receiver failure.
One method of increasing carrying capacity without incurring additional installation costs has been to multiplex multiple signals onto a single fiber using various methods, such as time division multiplexing, where two or more different signals are carried over the same fiber, each sharing a portion of time. Another more preferred multiplexing method is wavelength division multiplexing (WDM), where two or more different wavelengths of light are simultaneously carried over a common fiber.
Wavelength division multiplexing can separate a fiber's bandwidth into multiple channels. Dividing bandwidth into multiple discreet channels, such as 8, 16, 40, or even as many as 160 channels, through a technique referred to as dense channel wavelength division multiplexing (DWDM), is a relatively lower cost method of substantially increasing telecommunication capacity, using existing fiber optic transmission lines. Techniques and devices are required, however, for multiplexing the different discreet carrier wavelengths. That is, the individual optical signals must be combined onto a common fiber-optic line or other optical waveguide and then later separated again into the individual signals or channels at the opposite end or other point along the fiber-optic cable. Thus, the ability to effectively combine and then separate individual wavelengths (or wavelength sub-ranges) is of growing importance to the fiber-optic telecommunications field and other fields employing optical instruments.
Optical multiplexers are known for use in spectroscopic analysis equipment and for the combination or separation of optical signals in wavelength division multiplexed fiber-optic telecommunications systems. Known devices for this purpose have employed, for example, diffraction gratings, prisms and various types of fixed or tunable filters.
Approaches for selectively removing or tapping a channel, i.e., selective wavelengths, from a main trunk line carrying multiple channels, i.e., carrying optical signals on a plurality of wavelengths or wavelength sub-ranges, is suggested, for example, in U.S. Pat. No. 4,768,849 to Hicks, Jr. Hicks, shows filter taps, as well as the use of gangs of individual filter taps, each employing high performance, multi-cavity dielectric pass-band filters and lenses for sequentially removing a series of wavelength sub-ranges or channels from a main trunk line. The filter tap of Hicks, returns a multi-channel signal to the main trunk line as it passes the desired channel to a branch line. One known demux is disclosed in Pan et al., U.S. Pat. No. 5,652,814, FIG. 25. In Pan et al., the WDM input signal is cascaded through individual filter assemblies, consisting of a fiber collimator, thin film filter, and a fiber-focusing lens. Each filter is set for a given wavelength. However, aligning the fibers for each wavelength is costly and errors in the alignment contribute significantly to the system losses. Further, FIG. 13 of Pan et al. teaches the use of a dual fiber collimator, thin film filter, and a dual fiber focusing lens to selectively DROP and ADD a single wavelength or range of wavelengths. As discussed above, aligning the collimators is expensive.
Polarization dependent loss (PDL) is also a problem in WDM system because the polarization of the light drifts as it propagates through the fiber and furthermore this drift changes overtime. Thus, if there is PDL in any component, the drifting polarization will change the signal level, which may degrade the system operation.
Other multiplexer devices may be employed to add or drop channels in WDM systems. These systems are commonly known as optical add/drop multiplexers, or OADM. Another OADM, disclosed by Mizrahi in U.S. Pat. No. 6,185,023, employs fiber Bragg gratings to demux and mux signals in a WDM system. This method requires optical circulators and multiple components.
However, the multi channel OADM designs discussed above are not programmable by the end user. That is, each multiplexer is designed and manufactured to mux (add) specific channels; or when used in reverse each multiplexers is also designed and manufactured to demux (drop) specific channels. This limitation mandates that the optical system's parameters be fixed before installation. Changes are not possible without replacing the fixed optical multiplexers with different designed multiplexers. This is expensive.
One known programmable ROADM is discussed in Boisset et al, International Publication No. WO01/13151. In Boisset et al., the desired add/drop channel is programmed by translating a segmented filter. To achieve this translation however, a large mechanical mechanism is employed. A further limitation to Boisset et al. is that only a single channel may be added or dropped per device. Designers may employ multiple devices, deployed in series, and programmed as necessary to add/drop the correct channel; however, this approach requires multiple devices and has multiple points of failure. Furthermore, the size of such a device would be overly large and therefore not practical for many applications where space is limited. This ROADM is not East-West separable.
An OADM disclosed by Patel et al., U.S. Pat. No. 5,414,540 uses bulk gratings to demultiplex and multiplex WDM input and output signal and compact liquid crystal switches. Because the device uses polarization to switch the light path, the arbitrarily polarized incident beam must be converted into a singular polarization prior to switching by the liquid crystal. Patel teaches the use of a birefringent crystal and a Wollaston prism to separate the incident beam into two polarizations state located between the focusing lens and the liquid crystal. While the OADM disclosed by Patel is relatively compact; it only provides 2×2 switching for each wavelength. There is an Input and Add channel that may be selectively sent to either the Output or Drop channel. If higher dimensionality switching is required, then additional components are required. The additional components require additional space, add attenuation, and add cost to the system. A 2×2 switch has four sub beams incident on the liquid crystal (because of the conversion from an arbitrary polarized beam to a single polarization for the liquid crystal switch) and four sub beams leaving the liquid crystal. Thus, the aperture of the lens focusing the light on the grating must be a minimum of 4× larger than that required for a single sub beam in one polarization. This ROADM is not East-West separable.
An ROADM disclosed by Ranalli et al., U.S. Pat. No. 6,285,500, that uses bulk gratings to demultiplex and multiplex WDM input and output signal and compact liquid crystal switches. Because the device uses polarization to switch the light path, the arbitrarily polarized incident beam must be converted into a singular polarization prior to switching by the liquid crystal. Ranalli teaches the use of half-wave plates and a thin film polarization beamsplitter located before the lens that focuses light onto the liquid crystal. Because of the optical arrangement, the aperture of the lens focusing the light on the grating must be a minimum of 2× larger than that required for a single sub beam in one polarization. While the OADM disclosed by Ranalli is relatively compact; it only provides 2×2 switching for each wavelength. There is an Input and Add channel that may be selectively sent to either the Output or Drop channel. If higher dimensionality switching is required, then additional components are required. The additional components require additional space, add attenuation, and add cost to the system. This ROADM is not East-West separable.
A OADM disclosed by Patel et al., U.S. Pat. No. 6,327,019, uses bulk gratings to demultiplex and multiplex WDM input and output signal and compact liquid crystal switches. The OADM disclosed by Patel provides for dual 2×2 switching for each wavelength. There are two Input and two Add channels that may be selectively sent to either the two Output or two Drop channels. If higher dimensionality switching is required, then additional components are required. The additional components require additional space, add attenuation, and add cost to the system. Because liquid crystals use polarization to switch the light path, the arbitrarily polarized incident beam must be converted into a singular polarization prior to switching, which doubles the required aperture of the lens. Thus, the dual 2×2 switch has eight sub beams incident on the liquid crystal and eight sub beams leaving the liquid crystal. Thus, the aperture of the lens focusing the light on the grating must be a minimum of 8× larger than the aperture required for single incident beam in one polarization.
An ROADM disclosed by Aksyuk, et al, U.S. Pat. No. 6,204,946 uses a bulk grating to demultiplex and multiplex WDM input and output signal and Micro Electrical Mechanical Systems (MEMS) to provide the switching. This is another relatively compact switch, but it only provides 2×2 switching for each wavelength. There is an Input and Add channel that may be selectively sent to either the Output or Drop channel. If higher dimensionality switching is required, then additional components are required. The additional components require additional space, add attenuation, and add cost to the system. Because Aksyuk uses circulators to separate the Input and Add channels from the Output and Drop channels, the aperture of the lens focusing the light on the grating must be a minimum of 2× larger than the of a single incident beam. This ROADM is not East-West separable.
Another known programmable OADM is discussed Tomlinson, U.S. Pat. No. 5,960,133, uses a bulk gratings to demultiplex and multiplex WDM input and output signal, and MEMS mirrors to switch. The OADM disclosed by Tomlinson is programmable and provides for dual 2×2 switching. Tomlinson teaches a switch that does not require the use of circulators, potentially increasing the system efficiency. However, the aperture of the lens focusing the light on the grating must be a minimum of (1+Sqrt[2])× larger than the of a single incident beam for a 2×2 switch. Furthermore, for a dual 2×2 without circulators, the aperture of the lens focusing the light on the grating must be a minimum of Sqrt[10]× larger than that of a single incident beam. Thus, the size and expense of the focusing lens required grows quickly when moving from a single to dual switching. This ROADM is not East-West separable.
A programmable optical multiplexer/demultiplexer, disclosed by Marom et al, in U.S. Pat. App. Ser. No. 02/0196520, independently assigns every input optical channel in a signal to depart from any desired output port, which provides the functionality of 1×N switching for every wavelength. Marom teaches the use of a bulk grating to multiplex/demultiplex WDM input and output signal, and MEMS mirrors to switch. The demultiplexer device can also be operated in the reverse direction, and thus achieve programmable optical multiplexer functionality. However, the size and expense of the lens required by the demultiplexer also grows linearly with port count. A 1×5 port programmable optical multiplexer/demultiplexer requires a lens to focus light on the MEMs mirrors with an aperture at least 5× as large as that of a single incident beam.
Optical gratings are a periodic structure, which diffract light according to the wavelength. They can be used in either reflection or transmission. Gratings can be produce by modulating the surface height of a substrate or by modulating the index of refraction of a structure.
The spectral resolving power, R=λ/Δλ, of a grating is a measure of its ability to separate adjacent spectral lines, where λ is average wavelength of a line and Δλ is the limit of resolution. The theoretical resolving power is:R=N d cos Γ(sin α+sin β)/λwhere N is the number of groves, d is the groove spacing, β is the angle between the incident light path and the plane perpendicular to the groves, α is the angle of incidence on the grating and β is the angle of diffraction. If the grating is planar and the groove spacing is uniform, then the resolving power is proportional to the ruled with of the grating, N d. Spectral resolving power is an important design parameter; the greater the resolving power the greater the optical separation between channels, and ultimately the channels a grating-based system can accommodate. For low-loss transmission of OC-768 channels and a channel spacing of 100 GHz, it is preferred that the resolution be 20 GHz or finer.
Of course, a larger grating can be employed to increase the spectral resolving power, however, that requires a combination of more physical space and faster or longer focal length lenses that are more expensive. Another approach has been to decrease the spacing of the grating grooves, d. However, the maximum theoretical efficiency of the grating decreases for small groove separations. When the separations between the grooves spacing is comparable to the wavelength of light, it is possible to get gratings that operate with high efficiency (>90%) for any incident polarization state. As the groove spacing approaches half the wavelength of light, it is possible to get high efficiency for only light polarized parallel to the grooves. For even smaller grooves separations, it is not possible to get high efficiency in either polarization state. Thus, there is a practical limit to increasing spectral resolving power through decreased grating groove separations. The relationship between grating efficiency, polarization, and groove shape is well known in the art and described in Diffraction Grating Handbook, Ch. 9, 4th Ed, Richardson Grating Laboratory, C. Palmer, (2000), which is hereby incorporated by reference. Each bulk diffraction grating device requires a minimum number of grating grooves to achieve a given spectral resolution. The minimum size is determined by the optical configuration of the device and the grating parameters.
In cases where one or more channels are selectively removed or tapped at a node, as discussed above, channels that are not removed are often passed through the node and multiplexed back onto the main trunk line. One conventional method for passing a channel is to convert the channel sent in the optical domain to the electrical domain, and then to simply retransmit the signal onto the main trunk line at each node. Typically, the conversion required only allows the add/drop system to be operable for a particular protocol, at one particular speed, and for a certain number of wavelengths.
Several methods for optically directing passed through channels are currently practiced in East-West separable ROADM. Each node is connected to the main trunk line via two add/drop multiplexers, the first configured at the receiving side of the node, and the second at the transmitting side of the node. In one method all of the system's optical channels are first demultiplexed and each directed to a 1×2 optical switch. Depending on the programmed state of the 1×2 switch, the channels are either dropped or passed. Passed channels are directed to a 2×1 optical switch (the first input connects to the output of the 1×2 switch and the second is connected to an add port). The 2×1 optical switch is then connected to the add/drop multiplexer on the transmit side of the node. In this manner each channel that is transmitted onto the main trunk line (those that are passed as well as those that are added) are multiplexed by the add/drop multiplexer at the transmit side of the node. This method requires each add/drop multiplexer have as many ports as channels, and additionally requires two 1×2 switches for each channels. These additional components, and required ports make this method impractical for larger channel systems.
A second method currently practiced for optically directing passed through channels in an East-West separable ROADM connects each node to the main trunk line via two add/drop multiplexers, the first configured at the receiving side of the node, and the second at the transmitting side of the node. Waverka et-al teaches one such device in WO 01/37021, which is fully incorporated by reference herein. Passing of optical channels is achieved by first demultiplexing all the channels at the node's receiver-side add/drop multiplexer, then multiplexing the passed channels together at the receiver-side add/drop multiplexer and directing the multiplexed passed channels to the transmitter-side add/drop multiplexer. The passed channels are then demultiplexed, and any added channels are then multiplexed together with the passed channels and sent back onto the main truck line. While this method provides full programmability at the node and reduces the system components, the passed channels must be demultiplexed, multiplexed, demultiplexed, and multiplexed, thereby reducing the optical power of each channel.
The ability to switch to a number of optical ports in wavelength switches introduces another limiting design factor. In order to switch to a number of physical ports the size of the device must not only accommodate the space needed for the ports, but the optics must also direct the optical signals to those ports. As the number of ports increases the optical directing means (typically a moveable mirror) must be capable of directing the optical beams across a larger physical area where the optical ports are located. Also, as the optical beams must exit the ports within an acceptance angle so as to be coupled into the optical fiber, the ports must be physically located within a certain placement angle from the directing means. As the placement angle increases, the optical directing means generally becomes more expensive and the insertion loss increases. An additional lens may be used to focus the beam—however, this adds component cost and size to the device.
If the optical beams inside the device are made larger so as to increase spectral resolution the device size must increase, and in some cases larger lenses must be used. For example, an optical switch of the type disclosed by Marom et al. US 2002/0196520 A1, with one input port and four output ports (1×4) might be capable of switching 64 wavelengths spaced at 100 GHz. If the same design were used to switch 16 ports the grating and the grating aperture would likely need to be 4× larger to accommodate 100 GHz channels or if the grating was the same size, the system could switch 16 wavelength channels spaced at 400 GHz. The device disclosed by Marom cannot provide adequate spectral resolution for a large number of ports and a large number of wavelengths using small compact lenses that are easy to manufacture.
An optical wavelength switch disclosed by Waverka et al. WO 01/37021 uses a bulk diffraction grating and MEMS mirrors to provide 1×N switching. However, this design has a major drawback. Because the image is translated at the spectral focal plane by the MEMS mirrors, the incident angle on the grating changes with switch position, which in turn changes the angular dispersion provided by the grating. Thus, the device is unable to achieve adequate spectral resolution for a large number of ports and a large number of wavelengths with low losses. Waverka also teaches the use of cylindrical optics to produce an elliptical beam that minimizes the size of the grating. However, because the cylindrical optics are used symmetrically to both collimate light for the grating and to focus the light on the switch array, the footprint of the optical beam at the switch is a very high aspect ratio ellipse. Thus, very long thin, hard to fabricate switches are required.